Generate Workout Plans Directly in Chat

Build a Plan Without Leaving the Conversation:

You can ask Repley to generate a workout plan for you in chat — no need to open the Plan Generator wizard. The plan generates in the background while you keep chatting, and the result appears as an interactive widget with Save and Assign buttons built in.

How to start:

- Tap the Build new plan chip in the chat suggestions, OR - Just ask: "Help me create a workout plan for Lisa", "Build a 4-day strength plan for Sarah, full gym, intermediate", "Generate a HYROX plan for a new client"

What Repley will do:

1. Ask any quick clarifying questions it needs (goal, days per week, equipment, experience, injuries, target client) — read-the-room style. If your prompt is already specific, it skips straight to generation. 2. Resolve the client by UUID (privacy: it calls list_clients to look up the client first, never sends the name to the AI). 3. Kick off a background generation job. The chat keeps flowing — you can switch tabs, the message list stays interactive, and a small progress indicator floats at the bottom-right when the chat sidebar is closed.

Default first generation = 1 week. This is deliberate: 1 week generates in ~10-15 seconds so you see something to react to fast. The widget then offers an Expand to 4 weeks button which carries forward any edits you made and any preferences you mentioned in chat since.

The plan widget (when generation completes):

- Header — goal, weeks, days/week, and either the client's first name (for client-specific plans) or "template only" - Week-by-week summary — expandable per-week breakdown showing each day name with the top exercises and a count - Save as Template — saves directly to your library, no modal hop. Toast confirms with a "View" link to jump to the saved template. - Save + Assign to {client} — opens the assign modal pre-loaded with the generated plan, the client, and a default start date. Use this when you're ready to drop it onto a client's calendar in one continuous flow. - Pick a client to assign — when no client was named, opens the assign modal with a client picker so you can choose who it's for. - Expand to 4 weeks — only on 1-week first-pass plans. Triggers a second background generation with the original week 1 + your edits + chat context as continuation context. - Open full wizard — escape hatch that hands the chat-generated plan over to the Plan Generator wizard at the review step. Use this if you want the detailed builder UI (drag, swap exercises, manual sets/reps tweaks) instead of a Save/Assign-and-go.

Cost protection:

- Once a plan has been generated for a specific message, refreshing the chat or reopening the conversation will not regenerate it — the result is preserved. - Server-side request-hash dedup catches accidental double-clicks within 60 seconds, so a real double-fire collapses cleanly instead of running OpenAI twice. - Different parameters (different progression style, different experience level, etc.) within 60 seconds DO produce a fresh plan — only identical requests are deduplicated.

Save+Assign flow on /clients:

After tapping Save+Assign, you land on the Clients page with the assign modal already open and the plan pre-loaded. Confirm or change the start date, click Assign, and you're done. The plan is saved as a template AND assigned to the client in one step.

When to use chat vs. the wizard:

- Chat — you have a clear plan in mind ("4 days, full gym, knee-safe substitutes for Lisa") and want it generated and assigned in under a minute - Wizard — you want to step through every option (sport-specific configuration, weak HYROX stations, primary distances for sprint/endurance, advanced progression style)

Both paths use the same underlying AI generation engine and produce equivalent plans — the difference is the UI surface, not the output quality.

Important: AI-generated programs are a starting point, not a substitute for your professional judgement. Always review exercises, volume, and intensity for each client's individual needs, injury history, and training background before assigning a plan.